ayira-medium / README.md
predika-ai's picture
Update README.md
d283bb4 verified
metadata
library_name: transformers
language:
  - ht
license: apache-2.0
base_model: openai/whisper-medium
tags:
  - generated_from_trainer
datasets:
  - haitian-creole-asr
metrics:
  - wer
model-index:
  - name: Ayira Medium
    results:
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: Haitian Creole ASR Dataset
          type: haitian-creole-asr
          args: 'language: ht, split: train'
        metrics:
          - type: wer
            value: 0.1610558530986993
            name: Wer

Ayira Medium

This model is a fine-tuned version of openai/whisper-medium on the Haitian Creole ASR Dataset dataset. It achieves the following results on the evaluation set:

  • Loss: 0.2185
  • Wer: 0.1611

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 1e-05
  • train_batch_size: 4
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 16
  • optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 500
  • num_epochs: 10
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
0.1675 1.0 339 0.2434 0.2363
0.1356 2.0 678 0.2046 0.3384
0.0581 3.0 1017 0.1919 0.2433
0.0289 4.0 1356 0.1972 0.1909
0.0192 5.0 1695 0.2027 0.1787
0.005 6.0 2034 0.2043 0.1019
0.0036 7.0 2373 0.2093 0.1450
0.0006 8.0 2712 0.2147 0.1264
0.0004 9.0 3051 0.2162 0.1624
0.0003 10.0 3390 0.2185 0.1611

Framework versions

  • Transformers 4.52.4
  • Pytorch 2.6.0+cu124
  • Datasets 3.6.0
  • Tokenizers 0.21.1